In [1]:
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from warnings import filterwarnings
filterwarnings('ignore')
Reading dataset
In [3]:
df= pd.read_csv(r'C:\Users\Administrator\Documents\Data Analytics Real World Project -Python\3-Zomato Data Analysis/zomato.csv')
In [4]:
df.head()
Out[4]:
Exploratory data analysis
In [5]:
df.shape
Out[5]:
In [6]:
df.dtypes
Out[6]:
In [7]:
len(df['name'].unique())
Out[7]:
There are 8792 unique restaurants¶
In [10]:
#checking null values
df.isna().sum()
Out[10]:
Getting features with NAN values
In [12]:
features_na=[feature for feature in df.columns if df[feature].isna().sum() > 0]
features_na
Out[12]:
Checking the % of missing values
In [13]:
for feature in features_na:
print(' {} has {}% missing values '.format(feature,np.round(df[feature].isna().sum()/len(df)*100,4)))
In [15]:
df['rate'].unique()
Out[15]:
Dropping the NAN values
In [16]:
df.dropna(axis='index',inplace=True,subset=['rate'])
df['rate'].unique()
Out[16]:
In [18]:
df.shape
Out[18]:
Removing evrything after / from rate
In [19]:
def split(x):
return x.split('/')[0]
In [20]:
df['rate']=df['rate'].apply(split)
df['rate'].unique()
Out[20]:
Replacing NEW and - with 0
In [21]:
df.replace('NEW',0,inplace=True)
In [22]:
df.replace('-',0,inplace=True)
In [23]:
df['rate'].unique()
Out[23]:
changing the datatype of rate to float
In [24]:
df['rate']=df['rate'].astype(float)
In [25]:
df['rate'].dtype
Out[25]:
Calculating Avg rating of each restaurant¶
In [28]:
df.groupby('name')['rate'].mean().nlargest(20).plot.bar()
Out[28]:
In [35]:
df_rate=df.groupby('name')['rate'].mean().to_frame()
df_rate=df_rate.reset_index()
df_rate.columns=['Restaurant','Rating']
In [36]:
df_rate.head(20)
Out[36]:
In [38]:
df_rate.shape
Out[38]:
In [39]:
sns.set_style(style='whitegrid')
sns.distplot(df_rate['Rating'])
Out[39]:
50 % of the restaurants have rating between 3-4, While restaurants with rating above 4.5 are very rare.
In [ ]:
Finding the top restaurants in Bangaluru
In [41]:
plt.figure(figsize=(10,7))
chains=df['name'].value_counts()[0:20]
sns.barplot(x=chains,y=chains.index,palette='deep')
plt.title("Most famous restaurants chains in Bangaluru")
plt.xlabel("Number of outlets")
Out[41]:
In [49]:
fig1=px.bar(x=chains,y=chains.index)
fig1.show()
Checking how many Restaurants accept online orders
In [42]:
x= df['online_order'].value_counts()
labels=['Accepted','Not Accepted']
plt.pie(x,explode=[0.0,0.1],autopct='%1.1f%%')
Out[42]:
Plotting using plotly
In [43]:
import plotly.express as px
In [44]:
x= df['online_order'].value_counts()
labels=['Accepted','Not Accepted']
fig=px.pie(df,values=x,names=labels,title='Pie Chart for Online Order Acceptance')
In [45]:
fig.show()
Checking the ratio between the restaurants proving Table Booking
In [50]:
x= df['book_table'].value_counts()
labels=['Allowed','Not Allowed']
fig=px.pie(df,values=x,names=labels,title='Pie Chart for Online Booking tables')
fig.show()
Using Graph_obj in plotly
In [51]:
import plotly.graph_objs as go
from plotly.offline import iplot
In [52]:
x= df['book_table'].value_counts()
labels=['Allowed','Not Allowed']
In [59]:
trace=go.Pie(labels=labels,values=x,
hoverinfo='label+value',textinfo='percent',
textfont=dict(size=25),
pull=[0, 0, 0,0.2, 0]
)
In [60]:
iplot([trace])
Analyzing different types of restaurants
In [61]:
df['rest_type'].isna().sum()
Out[61]:
In [64]:
df['rest_type'].dropna(inplace=True)
In [65]:
df['rest_type'].isna().sum()
Out[65]:
In [67]:
df['rest_type'].describe()
Out[67]:
In [70]:
plt.figure(figsize=(20,12))
df['rest_type'].value_counts().nlargest(20).plot.bar(color='red')
plt.gcf().autofmt_xdate()
Ploting using plotly
In [78]:
trace1= go.Bar(
x=df['rest_type'].value_counts().nlargest(20).index,
y=df['rest_type'].value_counts().nlargest(20),
name='rest_type')
In [79]:
iplot([trace1])
Noticing from the graph Quick bite restaurants are most famous in bangalore as it is the tech capital of India
Finding the highest voted restaurant
In [81]:
df.groupby('name')['votes'].max().nlargest(20).plot.bar(color='red')
Out[81]:
Now using plotly
In [82]:
trace2=go.Bar(
x=df.groupby('name')['votes'].max().nlargest(20).index,
y=df.groupby('name')['votes'].max().nlargest(20),
name='name'
)
In [83]:
iplot([trace2])
From the above bar graph: Big brewski brewing company is the highest voted rest in bangalore
In [ ]:
Total rest at different locations
In [84]:
df.groupby('location')['name'].unique()
Out[84]:
In [85]:
restaurant=[]
location=[]
for key,location_df in df.groupby('location'):
location.append(key)
restaurant.append(len(location_df['name'].unique()))
In [86]:
df_total=pd.DataFrame(zip(location,restaurant))
df_total.columns=['location','restaurant']
df_total.set_index('location',inplace=True)
df_total.sort_values(by='restaurant').tail(10)
Out[86]:
In [87]:
df_total.sort_values(by='restaurant').tail(10).plot.bar(color='purple')
Out[87]:
In [95]:
trace3=go.Bar(
x=df_total['restaurant'].nlargest(10).index,
y=df_total['restaurant'].nlargest(10),
name='restaurant'
)
In [96]:
iplot([trace3])
In [97]:
df.isnull().sum()
Out[97]:
In [98]:
#Representing different type of restaurant data in Percentage
In [101]:
((df['rest_type'].value_counts()/len(df))*100).nlargest(10).plot.bar(color='red')
Out[101]:
Analyzing data on the basis of cuisines
In [104]:
cuisines=df['cuisines'].value_counts()[:10]
trace4=go.Bar(x=cuisines.index,
y=cuisines,
name='cuisines')
In [105]:
iplot([trace4])
From the above data it's clear that North Indian food rocks!! even in Bangaluru
In [ ]:
Finidng out cost for 2 people at different rest.
In [ ]:
In [106]:
len(df['approx_cost(for two people)'].value_counts())
Out[106]:
In [107]:
df['approx_cost(for two people)'].isna().sum()
Out[107]:
In [108]:
df.dropna(axis='index',subset=['approx_cost(for two people)'],inplace=True)
In [109]:
df['approx_cost(for two people)'].isna().sum()
Out[109]:
In [110]:
df['approx_cost(for two people)'].unique()
Out[110]:
In [112]:
df['approx_cost(for two people)'].dtype
Out[112]:
In [113]:
df['approx_cost(for two people)']=df['approx_cost(for two people)'].apply(lambda x: x.replace(',',''))
In [114]:
df['approx_cost(for two people)'].unique()
Out[114]:
In [115]:
df['approx_cost(for two people)']=df['approx_cost(for two people)'].astype(int)
In [116]:
df['approx_cost(for two people)'].dtype
Out[116]:
In [117]:
sns.distplot(df['approx_cost(for two people)'])
Out[117]:
In [118]:
# above graph shows that most of the restaurant's approx cost for 2 is nearly 1000
In [ ]:
Analyzing Cost vs Rating of restaurants
In [121]:
plt.figure(figsize=(12,8))
sns.scatterplot(x="rate",y='approx_cost(for two people)',hue='online_order',data=df)
plt.show()
Analyzing above graphs , we notice that most of the restaurants accepting online orders are higly rated and affordable
While most restaurants that doesn't accpet online orders are Higly rated but Expensive
In [ ]:
Finding difference between votes for restaurants accepting or not accepting online orders
In [127]:
sns.boxplot(x='online_order',y='votes',data=df)
Out[127]:
In [130]:
fig = px.box(df,x='online_order',y='votes')
fig.show()
from this boxplot,we can observe that median number of votes for both categories vary. Restaurants accepting online orders get more votes from customers as there is a rating option poping up after each order through zomato application.
In [ ]:
Price vs Online ordering
In [128]:
sns.boxplot(x='online_order',y='approx_cost(for two people)',data=df)
Out[128]:
In [129]:
fig = px.box(df,x='online_order',y='approx_cost(for two people)')
fig.show()
Restaurants with online orders are more affordable than the ones with no online ordering
In [ ]:
Cheapest cost restaurants for 2 people
In [133]:
df['approx_cost(for two people)'].min()
Out[133]:
In [137]:
df[df['approx_cost(for two people)']==40]
Out[137]:
most expensive cost for 2 people
In [135]:
df['approx_cost(for two people)'].max()
Out[135]:
In [136]:
df[df['approx_cost(for two people)']==6000]
Out[136]:
HistPlot for cost for 2 people
In [140]:
plt.figure(figsize=(15,9))
sns.histplot(df,x='approx_cost(for two people)')
plt.show()
In [141]:
px.histogram(df, x="approx_cost(for two people)")
Mostly cost of all restaurants is below Rs.1000 but some are expensive
In [ ]:
Most costly Rate for 2 people is served at which Restaurant what exactly is the dish involved in this and liked dish of that restaurant
In [142]:
df[df['approx_cost(for two people)']==6000].loc[:,('name','cuisines','dish_liked')]
Out[142]:
In [143]:
df[df['approx_cost(for two people)']==6000][['name','cuisines','dish_liked']]
Out[143]:
In [144]:
data=df.copy()
In [145]:
data.dtypes
Out[145]:
In [147]:
data.set_index('name',inplace=True)
In [148]:
### Top 10 Most Expensive restaurant with approx cost for 2 people
In [149]:
data['approx_cost(for two people)'].nlargest(10).plot.bar()
Out[149]:
In [150]:
trace1=go.Bar(
x=data['approx_cost(for two people)'].nlargest(10).index,
y=data['approx_cost(for two people)'].nlargest(10),
name='Priority'
)
In [151]:
iplot([trace1])
Top 10 Cheapest restaurant with approx cost for 2 people
In [152]:
data['approx_cost(for two people)'].nsmallest(10).plot.bar()
Out[152]:
In [153]:
trace1=go.Bar(
x=data['approx_cost(for two people)'].nsmallest(10).index,
y=data['approx_cost(for two people)'].nsmallest(10),
name='Priority'
)
In [154]:
iplot([trace1])
Cheapest restaurants location wise
In [164]:
#data.set_index('location',inplace=True)
data['approx_cost(for two people)'].nsmallest(20).plot.bar()
Out[164]:
In [166]:
data['approx_cost(for two people)'].nlargest(20).plot.bar()
Out[166]:
In [ ]:
Finding Restaurants having rating above 4 and affordable
In [169]:
df[(df['rate']>=4) & (df['approx_cost(for two people)']<=500)].shape
Out[169]:
In [170]:
df_new=df[(df['rate']>=4) & (df['approx_cost(for two people)']<=500)]
In [171]:
len(df_new['name'].unique())
Out[171]:
Finding total such restaurants at different locations
In [172]:
location=[]
total=[]
for loc,location_df in df_new.groupby('location'):
location.append(loc)
total.append(len(location_df['name'].unique()))
In [173]:
len(location)
Out[173]:
In [174]:
len(total)
Out[174]:
In [180]:
location_df=pd.DataFrame(zip(location,total))
location_df.columns=['location','restaurants']
location_df.set_index('location',inplace=True)
In [182]:
location_df
Out[182]:
In [189]:
location_df['restaurants'].nlargest(10).plot.bar()
plt.gcf().autofmt_xdate()
plt.ylabel('Total restaurants')
Out[189]:
In [190]:
trace1 = go.Bar(
x = location_df['restaurants'].nlargest(10).index,
y = location_df['restaurants'].nlargest(10),
name= 'Priority')
iplot([trace1])
Finding best budget restaurant in any location
Creating a function which takes restaurant type and location and return names .
In [191]:
def return_budget(location,restaurant):
budget=df[(df['approx_cost(for two people)']<=400) & (df['location']==location) &
(df['rate']>4) & (df['rest_type']==restaurant)]
return(budget['name'].unique())
In [192]:
return_budget('BTM',"Quick Bites")
Out[192]:
In [ ]:
Finding areas famous for restaurants
In [193]:
plt.figure(figsize=(10,7))
Restaurant_locations=df['location'].value_counts()[:20]
sns.barplot(Restaurant_locations,Restaurant_locations.index)
Out[193]:
In [194]:
Restaurant_locations=df['location'].value_counts()[:20]
trace1 = go.Bar(
x = Restaurant_locations.index,
y = Restaurant_locations,
name= 'Priority')
iplot([trace1])
In [ ]:
Performing Geographical Analysis
To do so we need longitudes and lattitudes of any location, hence we will use Geopy
In [195]:
df.shape
Out[195]:
In [196]:
len(df['location'].unique())
Out[196]:
In [197]:
locations=pd.DataFrame({"Name":df['location'].unique()})
In [200]:
locations['new_Name']='Bangalore '+locations['Name'] #Adding bangalore to the name
In [201]:
locations.head()
Out[201]:
In [202]:
!pip install geopy
In [203]:
from geopy.geocoders import Nominatim
In [204]:
lat_lon=[]
geolocator=Nominatim(user_agent="app")
for location in locations['Name']:
location = geolocator.geocode(location)
if location is None:
lat_lon.append(np.nan)
else:
geo=(location.latitude,location.longitude)
lat_lon.append(geo)
In [205]:
locations['geo_loc']=lat_lon
In [206]:
locations.head()
Out[206]:
In [207]:
locations.to_csv('zomato_locations.csv',index=False)
In [ ]:
Now we can use these locations to plot maps
In [208]:
Rest_locations=pd.DataFrame(df['location'].value_counts().reset_index())
In [209]:
Rest_locations.columns=['Name','count']
Rest_locations.head()
Out[209]:
In [210]:
locations.head()
Out[210]:
In [211]:
locations.shape
Out[211]:
In [212]:
Rest_locations.shape
Out[212]:
In [213]:
Restaurant_locations=Rest_locations.merge(locations,on='Name',how="left").dropna()
Restaurant_locations.head()
Out[213]:
In [214]:
Restaurant_locations.shape
Out[214]:
In [215]:
Restaurant_locations['count'].max()
Out[215]:
In [216]:
type(Restaurant_locations['geo_loc'][0])
Out[216]:
In [217]:
def generateBaseMap(default_location=[12.97, 77.59], default_zoom_start=12):
base_map = folium.Map(location=default_location, zoom_start=default_zoom_start)
return base_map
In [218]:
len(Restaurant_locations['geo_loc'])
Out[218]:
In [219]:
Restaurant_locations.isna().sum()
Out[219]:
In [220]:
Restaurant_locations['geo_loc'][0][0]
Out[220]:
In [221]:
Restaurant_locations['geo_loc'][0][1]
Out[221]:
In [222]:
np.array(Restaurant_locations['geo_loc'])
Out[222]:
In [223]:
#### unzip it
lat,lon=zip(*np.array(Restaurant_locations['geo_loc']))
In [224]:
type(lat)
Out[224]:
In [225]:
Restaurant_locations['lat']=lat
Restaurant_locations['lon']=lon
In [226]:
Restaurant_locations.head()
Out[226]:
In [227]:
!pip install folium
In [228]:
import folium
from folium.plugins import HeatMap
basemap=generateBaseMap()
In [229]:
basemap
Out[229]:
In [230]:
Restaurant_locations[['lat','lon','count']].values.tolist()
Out[230]:
In [231]:
HeatMap(Restaurant_locations[['lat','lon','count']].values.tolist(),zoom=20,radius=15).add_to(basemap)
Out[231]:
In [232]:
basemap
Out[232]:
Most of the restaurants are in Central Bangalore area as compared to other areas. Potenital Entrepreneurs can refer an find a suitable location for new restaurants.
In [ ]:
Heatmap of north indian restaurants
In [233]:
df2= df[df['cuisines']=='North Indian']
df2.head()
Out[233]:
In [234]:
north_india=df2.groupby(['location'],as_index=False)['url'].agg('count')
north_india.columns=['Name','count']
north_india.head()
Out[234]:
In [235]:
north_india=north_india.merge(locations,on="Name",how='left').dropna()
In [236]:
north_india['lan'],north_india['lon']=zip(*north_india['geo_loc'].values)
In [237]:
north_india.drop(['geo_loc'],axis=1)
Out[237]:
In [238]:
basemap=generateBaseMap()
HeatMap(north_india[['lan','lon','count']].values.tolist(),zoom=20,radius=15).add_to(basemap)
basemap
Out[238]:
In [ ]:
Analying reviews of particular restaurants
In [239]:
df.head()
Out[239]:
In [240]:
df['reviews_list'][0]
Out[240]:
In [241]:
data=df['reviews_list'][0].lower()
data
Out[241]:
In [242]:
import re
data2=re.sub('[^a-zA-Z]', ' ',data)
data2
Out[242]:
In [243]:
data3=re.sub('rated', ' ',data2)
data3
Out[243]:
In [244]:
data4=re.sub('x',' ',data3)
data4
Out[244]:
In [245]:
re.sub(' +',' ',data4)
Out[245]:
In [246]:
dataset=df[df['rest_type']=='Quick Bites']
In [247]:
type(dataset['reviews_list'][3])
Out[247]:
In [248]:
total_review=' '
for review in dataset['reviews_list']:
review=review.lower()
review=re.sub('[^a-zA-Z]', ' ',review)
review=re.sub('rated', ' ',review)
review=re.sub('x',' ',review)
review=re.sub(' +',' ',review)
total_review=total_review + str(review)
In [251]:
from wordcloud import WordCloud, STOPWORDS
stopwords=set(STOPWORDS)
wordcloud = WordCloud(width = 800, height = 800,
background_color ='white',
stopwords = stopwords,
min_font_size = 10).generate(total_review)
# plot the WordCloud image
plt.figure(figsize = (8, 8))
plt.imshow(wordcloud)
plt.axis("off")
Out[251]:
In [252]:
def importance(restaurant):
dataset=df[df['rest_type']==restaurant]
total_review=' '
for review in dataset['reviews_list']:
review=review.lower()
review=re.sub('[^a-zA-Z]', ' ',review)
review=re.sub('rated', ' ',review)
review=re.sub('x',' ',review)
review=re.sub(' +',' ',review)
total_review=total_review + str(review)
wordcloud = WordCloud(width = 800, height = 800,
background_color ='white',
stopwords = set(STOPWORDS),
min_font_size = 10).generate(total_review)
# plot the WordCloud image
plt.figure(figsize = (8, 8))
plt.imshow(wordcloud)
plt.axis("off")
In [253]:
importance('Quick Bites')
In [ ]: